I wonder how they will stand up against requests from the swiss government regarding lawful intercept access. Which, for larger providers is mandatory to participate in.
In true end-to-end encryption, this would probably not matter, since you can hand over all the encrypted e-mails you want and no one's going to be reading them unless they have your private keys. That said, the nature of in-browser crypto is such that they (or anyone who controls their servers) could intermittently change the JS code they are serving in such a way that it captures your private keys and decrypt all your e-mails.
So it really depends on your threat model. This service is somewhat more secure than Lavabit, but incrementally and not by leaps and bounds. It also constrains the attack model (in the Lavabit model they could be coerced to give the plaintext directly, in this case they would need to be coerced to actively steal their users' private keys).
There is no true E2E if you run it inside a browser.
And even if it would be application based (PGP, S/MIME), it would still leak metadata like crazy.
With all the threat models, I come to the conclusion, that there is no real E2E possible _at_all. All known platforms have been compromised, either by lawful interception/state trojan means or by direct hacking.
So it really depends on your threat model. This service is somewhat more secure than Lavabit, but incrementally and not by leaps and bounds. It also constrains the attack model (in the Lavabit model they could be coerced to give the plaintext directly, in this case they would need to be coerced to actively steal their users' private keys).